Every creator with an old video that stopped getting views eventually asks the same thing: would a better thumbnail bring it back? The advice online splits cleanly in half, and almost none of it shows its working.
So I went looking for cases where somebody actually measured it. Vevo ran the swap on 3,972 old videos at once. Ali Abdaal ran a controlled test on one. Ordinary creators on a forum reported it going both ways, sometimes the same person a year apart. Where a claim had a picture attached, I pulled the archived images and measured them myself instead of repeating the description.
By the end you will know which of your videos are worth re-thumbnailing, which ones cannot be saved by any image, and the exact check to run in YouTube Studio before you open a design tool.
Quick answer
- Vevo re-thumbnailed 3,972 old catalogue videos and saw a 12 percent average view lift. The best case, Halsey's "Ghost," jumped 4,000 percent in two weeks.
- Ali Abdaal reported one video going from 300,000 to 1,100,000 views after an A/B tested swap.
- Real creators report mixed results. The same person, on the same forum, saw it work once and fail a year later.
- The deciding factor is impressions, meaning whether YouTube is still showing the video to anyone. A new thumbnail cannot revive a video nobody sees.
- What Vevo's 3,972-Video Thumbnail Test Found
- Which Thumbnail Statistics Are Real And Which Are Just Copied
- What Creators Report After Changing An Old Thumbnail
- When Changing A Thumbnail Increases Views, And When It Does Not
- How To Change A Thumbnail Without Losing Views
- How A Thumbnail Change Works From The Viewer's Perspective
- Conclusion (What I Would Do)
- FAQ: Changing A YouTube Thumbnail
What Vevo's 3,972-Video Thumbnail Test Found
In April 2019, Vevo re-thumbnailed 3,972 old catalogue videos on YouTube, all of them live for years already.
That matters, because an old video and a fresh upload are different situations. A new upload has no track record. An old one already has one, for better or worse. Vevo tested the second case, which is the one most people asking this question are actually in.
The Four Cases At A Glance
- Vevo, 3,972 old videos: 12 percent average view lift in the first 20 days after each swap.
- Halsey, "Ghost (Room 93 Version)": dark image swapped for a brighter one. 4,000 percent increase in two weeks. I checked the archived images myself and this one holds up.
- Ali Abdaal: an A/B tested swap, 300,000 to 1,100,000 views.
- Tim Gabe, via Paddy Galloway: a reported 40x jump in daily views. Secondhand, no primary source, so treat it as a story.
Vevo's Numbers In Full
The average across the full sample was a 12 percent view lift in the first 20 days. It flattened after that but stayed above where each video started.
Twelve percent is not dramatic. It is real though, measured across nearly four thousand videos, which is more evidence than most articles on this topic bother to cite.
The Halsey Case, Checked Against The Actual Images
Inside that batch sat one outlier. Halsey's "Ghost (Room 93 Version)," already about four years old, gained 4,000 percent more views in the two weeks after its thumbnail changed. That is roughly forty times what it had been getting.
Vevo's VP, JP Evangelista, explained the change in one line.
"The thumbnail wasn't standing out, it wasn't popping."
The video is Halsey - Ghost (Room 93 Version), and you can open it now and see for yourself what it carries today.
Rather than repeat Vevo's description, I pulled the archived images. Here is the last pre-swap thumbnail the Wayback Machine captured, from 6 June 2018.
And here is what replaced it, archived 9 December 2018.
The pre-swap image measures 18.1 out of 255 for brightness. The post-swap one measures 113.7, more than six times brighter. That matches the reporting almost exactly: dark and low contrast out, brighter with two faces filling the frame in. The Wayback index shows the file changed between those two dates, inside Vevo's stated campaign window.
This Video Has Been Re-Thumbnailed Twice, Not Once
Here is the part that surprised me. That bright 2018 image is not what the video carries today.
Somewhere between December 2018 and now, Vevo changed it a second time, back to a dark night shot. The most recent Wayback capture, from 15 August 2026, measures 15.5 for brightness. That is almost exactly where it started before the 2019 campaign.
So the sequence on this one video is: dark (2018) to bright (the swap that produced the 4,000 percent jump) and back to dark (today).
I have no view or CTR data for that second change, and Vevo has never said anything about it publicly, so I am not going to guess why it happened. Open the video and you can check the current image yourself.
Worth sitting with though: the exact same team that ran the most-cited thumbnail experiment on the internet later walked its most successful result back. Nobody outside Vevo knows whether that was a brand decision, a new campaign, or a swap that quietly underperformed.
Ali Abdaal And Tim Gabe's Reported Results
Two more swaps round out the picture, one solid and one shaky. An A/B test, for reference, just means running two thumbnails against each other and keeping whichever more people click.
- Ali Abdaal: 300,000 to 1,100,000 views after a deliberate A/B test on 24 November 2023. I could not confirm the video's age, so this may not be an old-video case.
- Tim Gabe, via Paddy Galloway: a 40x jump in daily views off a thumbnail roughly 30 to 40 percent better. Secondhand, no primary post to check.
Vevo's catalogue and Ali Abdaal's video were both still being served to people when their thumbnails changed. Tim Gabe's is the one I cannot check either way, and that gap matters later.
Which Thumbnail Statistics Are Real And Which Are Just Copied
Before citing more numbers, here is how I sorted checkable evidence from figures nobody can trace to a source.
Where The 37 To 110 Percent Figure Actually Comes From
You will see a "37 percent to 110 percent CTR improvement" figure on a dozen SEO blogs about this exact topic. CTR means click-through rate, the share of people who saw your thumbnail and clicked it.
As far back as I could trace, it starts at TubeBuddy's own blog post on leaked YouTube A/B testing data. It has been copied from blog to blog since, with nobody independently checking it.
The Case Study That Never Shows A Number
1of10's post on updating old thumbnails uses Ryan Trahan's "I Fixed 100 Famous Logos" as its case study, and gives no view number at all. Just "a significant surge." A case study with no number is an anecdote wearing a suit.
The post does not show the thumbnails either, so I went looking in the Wayback Machine. These are the three captures I was able to find for that video. I am showing them as what I found, not as a complete history, because I cannot rule out other versions that were never archived.
Different brand, different colour, different tweet text. Whatever else is true about the view numbers, these are not small tweaks to one design. They are three genuinely different images on one video.
I have no view figure attached to any single swap, so I am not claiming this proves anything. I wanted to share it because it is the clearest example I could find of what a real thumbnail change actually looks like, as opposed to the phrase "we changed the thumbnail" doing all the work in somebody else's case study.
What I Trust And What I Do Not
| Confidence | Source |
|---|---|
| Trust | Vevo's Variety-reported numbers, and Ali Abdaal's own tweet |
| Checked myself | The Halsey thumbnails, measured off the Wayback captures |
| Not trusted | The "37 to 110 percent" CTR figure, since nobody traces it to a real test |
The Gap In Our Own Data, And Why It Exists
To be straight about the limits of this study: we could not produce this evidence ourselves either.
What would settle the question completely is one video, two thumbnails, at two known dates, with the view and impression numbers for each period sitting next to them. Nobody outside that channel can get those numbers. YouTube shows impressions and click-through rate only to the channel owner, inside their own Studio.
So everything published about thumbnail swaps, this study included, is either self-reported by the creator or reconstructed from the outside the way I did with the Halsey captures. That is worth knowing before you trust any number on this topic, including the ones here.
Get the next one Every Sunday we take one channel that just broke out and decode what actually changed. Free, one email. Subscribe here.
What Creators Report After Changing An Old Thumbnail
Big brands are not who is reading this, so here is the counterweight: ordinary creators on yttalk.com, a long-running forum, quoted exactly as written.
The Same Person, Two Opposite Results A Year Apart
In October 2016, a user called EVO posted this.
"I went back and changed a load of old video thumbnails (to improved thumbnails) and it increased viewing."
Then in September 2017, the same user posted this in a different thread.
"I've had to change thumbnails before and it has killed my vids."
Same person, opposite conclusion, about a year apart. That is the single most useful thing I found while studying this, because it kills the idea that there is one clean answer.
It also rhymes with something bigger. Vevo ran the most cited thumbnail experiment on the internet, and then quietly put a dark thumbnail back on its best result. I cannot tell you those two things are connected, because neither EVO nor Vevo published their numbers. But a solo creator and a company with a data team both ended up walking a swap back, and that is at least worth noticing.
When It Went Wrong
A user called borisv reported a straight negative with a number attached, on 4 October 2017. He replaced all of his thumbnails at once and it "affected my youtube suggestion in a negative way." Suggested traffic means the videos YouTube puts next to whatever someone is already watching.
His daily visits fell from around 16,000 to 7,000 and were still falling when he posted.
One detail matters more than the number: he changed everything at once, so he could never know which change did what.
When It Went Right
UhoohExtra, posting in September 2018 about testing on high traffic videos, took the opposite position.
"Changing the thumbnail won't hurt you now. I've tested on high traffic videos ... if you up your click through rate % with a better thumbnail then youtube will promote your video more."
These people are not lying about their own channels. Something else has to explain why the same action produced opposite outcomes, and that something is next.
When Changing A Thumbnail Increases Views, And When It Does Not
How YouTube's Own Creator Liaison Frames Repackaging
Rene Ritchie, YouTube's Creator Liaison, is about as close to a primary source as this topic gets.
"I change the thumbnail, change the title, and now I'm targeting the broader audience."
The key point is easy to miss. The algorithm is not rewarding you for the edit. It is reacting to how people respond to the new packaging.
Why Impressions Decide Everything
It comes down to one term: impressions, the number of times YouTube actually put your thumbnail in front of somebody.
If the algorithm reacts to how people respond, then a new thumbnail needs an audience in front of it to respond at all. No impressions means nobody sees the new image, so there is nothing to react to. The algorithm does not read your thumbnail the way a person does. It reads the clicks and watch time your thumbnail produces.
That reconciles most of it. Vevo's catalogue, Ali Abdaal's video and UhoohExtra's high traffic videos were all still being served when the thumbnail changed. Tim Gabe's case is the exception, and there is no way to check whether his video was still getting impressions.
TubeRanker's position backs this up, and also kills a common fear: changing a thumbnail does not reset a video's views, watch time or ranking.
The Decision Framework, In Three Cases
| Your video's situation | What to do |
|---|---|
| Still getting impressions, CTR below your channel average | This is the case worth trying |
| Still getting impressions, CTR at or above your channel average | The thumbnail is not holding it back. Leave it alone |
| Near zero impressions, genuinely dead | A new thumbnail has nothing to work with. This is the case most people asking are actually in |
The Ceiling Even A Good Thumbnail Has
OverseerOS names the limit well.
"A new thumbnail may make the value clearer. It cannot create value the video does not contain."
How To Change A Thumbnail Without Losing Views
All of this is doable this week, with nothing paid.
The 28-Day Check
- Open YouTube Studio, go to Analytics, then the Reach tab.
- Check the video's impressions and impressions click-through rate over the last 28 days.
- Compare that against your own channel's average, not a figure from a blog.
YouTube's own help page confirms this path.
There is no universal "good CTR," only your channel's normal.
Test On Your Second-Best Video First
Ezrider92356, posting on 24 September 2018, updated the thumbnail on his second most popular video as a test before touching his top performer.
Your best video is the worst place to experiment. And borisv is the cautionary tale for changing everything at once.
What To Actually Change In The Image
Vevo's team described their change as dark and low contrast out, brighter with bigger faces in. I confirmed that difference against the archived images earlier.
Ed Lawrence of Film Booth put the principle this way.
"simplify your message down to around four words basically your thumbnail needs to work when it's displayed very small"
Our own YouTube outlier breakdown on creator Toby backs this up. His outlier video keeps the face plain and the text to exactly three words: "ruined my business."
MrBeast made a narrower claim about his own channel.
"I closed my mouth on all my thumbnails and the watch time went up on every video lol."
Two caveats: that is his channel specifically, not a rule to copy, and I could not find a dated MrBeast video whose thumbnail I could confirm and show.
One more example of how far a swap can go. YouTube's own first video, "Me at the zoo," carried a plain dim photo for years.
In December 2023 it briefly carried this instead.
Both images look smaller than the others here because that is the real ceiling, not a crop. The video is from 2005 and YouTube never generated a maxresdefault image for it. This was a prank, reverted within about two weeks. I am not holding it up as a tactic, only as proof that a thumbnail can change completely whenever somebody wants it to.
YouTube's Own Test And Compare Tool
YouTube's native thumbnail test beats any manual swap, and it is not a limited beta any more. You need advanced features enabled on your channel, and you have to set it up on desktop in YouTube Studio. Creator Insider describes it as a way to "compare the performance of up to three thumbnails and select a winning thumbnail."
YouTube's help page confirms the mechanics: up to 3 thumbnails and titles, and whichever combination earns the highest watch time is shown to everyone afterward.
The same page says a test should finish within two weeks, close to the window Vevo and Halsey's numbers were measured over. One limit worth knowing before you plan around it: you cannot run this on Shorts, scheduled lives or premieres.
That is still a controlled test on your own traffic, which beats a guess based on somebody else's case study every time.
What A Real Revert Looks Like
"It worked" means impressions and CTR both moving, not one day of extra views.
If impressions fall the way borisv's did, put the old thumbnail back. Nothing about the video gets reset, so a revert costs only the time.
One limitation stands regardless: without the native test, a manual swap is a before and after, not a controlled experiment. Other things on your channel keep changing too.
How A Thumbnail Change Works From The Viewer's Perspective
One thing almost nobody discusses, and it is the part I keep coming back to.
When you change a thumbnail and a title, the video stops looking like something your audience has already seen. To a subscriber scrolling their feed, it reads as new. So they click.
The click is the easy part. What matters is what happens in the ten seconds after it, and that splits into three different journeys.
- Never seen it, watches it properly. The thumbnail did its job on somebody the video never reached the first time. They have no memory of it, so nothing interrupts them. This is a clean full view, and it is where the extra watch time in every case study above actually comes from.
- Seen it, clicked because the thumbnail looked new, backs out. This is the one that hurts. A few seconds in they recognise the intro and hit back. YouTube does not see "returning viewer being polite." It sees a click followed by an immediate exit, which is the same shape as a bad video. Stack enough of those and you get borisv's chart.
- Seen it, watches it again anyway. Loyal viewers who do not mind a rewatch. They deliver a full view on content they have already consumed, which is the best possible outcome and also the smallest group. They are the reason a swap rarely goes as badly as the worst forum stories suggest.
Group one is the upside, group three is a bonus, and group two is the entire risk. Every bad outcome in this study traces back to a swap that put too many people into group two.
Which group dominates is not luck. It is decided before you touch anything, by how many people already watched the video.
A video most of your subscribers already saw has a large group two sitting there waiting to bounce. A video that quietly underperformed on release barely has a group two at all, because hardly anyone saw it the first time. That is the real reason underperformers are the safe place to experiment, and your best video is the dangerous one. It is not superstition, it is just where your audience already is.
So my recommendation is to change it, but not blind. Run it through Test and Compare where you can, and only after the video clears the impressions check above. The rules in this study exist precisely so you are not gambling with the videos that are still working for you.
Conclusion (What I Would Do)
Changing a thumbnail can increase views on an old video, and there is real sourced evidence of it at both huge and small scale. But it only has a route to work while the video is still getting impressions. On a genuinely dead video it is mostly wasted effort. Done as a batch across many videos at once, it can go backwards.
Here is what I would personally do, in order:
- Check impressions first, before opening a design tool.
- Only touch videos still being served, where CTR sits under your channel average.
- Change one video, not twenty.
- Start with your second best video, not your best one.
- Use Test and Compare if your channel has advanced features enabled, rather than swapping blind.
- Give it about three weeks before deciding anything.
- Revert without ego if impressions fall.
The bigger point: a better thumbnail can win back clicks a video was already losing. It cannot make an old video interesting to people who were never going to want it.
If you have swapped a thumbnail on an old video, tell me what happened to your impressions and CTR. Specifically whether the video was still getting impressions before you touched it, because that is the one variable almost nobody reports.
FAQ: Changing A YouTube Thumbnail
Does changing your thumbnail reset your views or ranking?
No. TubeRanker's research says plainly that changing a thumbnail does not reset a video's views, watch time or ranking, and I could not find anything contradicting it.
How long should you wait before judging a thumbnail swap?
Go by the windows real people measured over. Vevo measured across the first 20 days. The Halsey jump was measured over two weeks. YouTube's own Test and Compare usually resolves inside two weeks as well. So give it two to four weeks, and do not judge it on day two.
Will changing a thumbnail hurt a video that is already doing well?
Only if you guess instead of checking. If your CTR is already at or above your channel average, the thumbnail is not what is holding the video back, and swapping trades a working image for an unknown one. Test on a smaller video first.
Does a thumbnail swap affect YouTube Shorts the same way?
I could not find a single source in this research that addresses Shorts specifically. Everything studied here is long-form. I am not going to guess whether the same impressions logic applies. Treat it as an open question.
What is YouTube's own thumbnail testing tool?
Test and Compare lets you run up to three thumbnails against each other on your own traffic. Whichever earns the highest watch time share is shown to everyone afterward. It usually resolves within two weeks.
Should I change many thumbnails at once?
No. borisv changed all of his and watched daily visits fall from 16,000 to 7,000 with no way of knowing which change caused it. Change one video at a time.